Innovating Risk Management and Hedging Strategy for Convertible Bonds Using Support Vector Machine

Abstract

Convertible bonds (CB) contain many kinds of embedded options and the complexity of their interaction makes hedging exposures of CBs challengeable. In order to tackle the issue, this paper introduced support vector machine (SVM) approach to overcome the shortcomings of traditional pricing methods and enhance hedging efficiency. By feature selection, kernel function determination and parameter optimization, SVM-based model proved to be more effective in that it can deal with complicated interaction among options of CBs and nonlinear and time-varying correlation among variant variables. On the basis of the proposed model, an innovated hedging strategy for CBs, based on delta-gamma-neutral hedging methodology, was explored in terms of the sensitivity of CB's value to the underlying stock price. Moreover, the model brought out great flexibility for risk management and hedge ratio determination by coping with neatly stochastic changes in volatility of the underlying stock, with remarkable advantages in hedging performance in empirical analysis over the traditional methods.